Synchronizing local compiler cache.

.gitattributes

@@ -7843,3 +7843,5 @@ neuronxcc-2.17.194.0+d312836f/MODULE_b74e4b6342153be8fdd2/vae_decoder/model.neur
 neuronxcc-2.17.194.0+d312836f/MODULE_b74e4b6342153be8fdd2/vae_encoder/model.neuron filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.17.194.0+d312836f/MODULE_0f0be0cf9fb35688b98c+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.17.194.0+d312836f/MODULE_45dec79bcb55d7eacd60+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.17.194.0+d312836f/MODULE_8bdbf5372d7324466d56+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.17.194.0+d312836f/MODULE_a64f774f616f6a5de9a2+613edded/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.17.194.0+d312836f/0_REGISTRY/0.2.0.dev2/inference/llama/deepseek-ai/DeepSeek-R1-Distill-Llama-8B/b8d797840e56152f0045.json

@@ -0,0 +1,45 @@
+{
+  "_entry_class": "SingleModelCacheEntry",
+  "_model_id": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
+  "_task": "text-generation",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "neuron": {
+    "auto_cast_type": "bf16",
+    "batch_size": 32,
+    "checkpoint_id": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
+    "checkpoint_revision": "6a6f4aa4197940add57724a7707d069478df56b1",
+    "compiler_type": "neuronx-cc",
+    "compiler_version": "2.17.194.0+d312836f",
+    "num_cores": 8,
+    "sequence_length": 4096,
+    "task": "text-generation"
+  },
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 8.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000.0,
+  "tie_word_embeddings": false,
+  "use_cache": true,
+  "vocab_size": 128256
+}

neuronxcc-2.17.194.0+d312836f/MODULE_8bdbf5372d7324466d56+613edded/compile_flags.json

@@ -0,0 +1 @@
+["--target=trn1", "--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.17.194.0+d312836f/MODULE_8bdbf5372d7324466d56+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a8ccb7ef6021b59665c8ecebec7ca80bc85d3804074b5c6a827e3b74aa83458
+size 363567

neuronxcc-2.17.194.0+d312836f/MODULE_8bdbf5372d7324466d56+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d930a32038dbe72596e2f7ab41af1a07aaa01ea32c303a57d4ed4754d7bbd83
+size 6595584

neuronxcc-2.17.194.0+d312836f/MODULE_a64f774f616f6a5de9a2+613edded/compile_flags.json

@@ -0,0 +1 @@
+["--target=trn1", "--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.17.194.0+d312836f/MODULE_a64f774f616f6a5de9a2+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40bed53b3250ff7ad4953eacf3d8cc807c81e465e3102998f0f857930961ba2d
+size 466825

neuronxcc-2.17.194.0+d312836f/MODULE_a64f774f616f6a5de9a2+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4c62baa4a7f0b2a93f53810d347bbef25843a89c49f1449092581408aaff6a9
+size 30628864